You are on page 1of 46

HSE

Health & Safety


Executive

Re-evaluation of fatigue curves for


flush-ground girth welds

Prepared by TWI Limited


for the Health and Safety Executive 2003

RESEARCH REPORT 090


HSE
Health & Safety
Executive

Re-evaluation of fatigue curves for


flush-ground girth welds

Dr Y-H Zhang, Dr S J Maddox


and Dr R Razmjoo
TWI Limited
Granta Park
Abington
Cambridge
CB1 6AL

The fatigue performance of girth welded steel pipes with the weld cap and root flush-ground is
designated as Class C in the BS 7608 code for fatigue design. However, this is not based on
experimental data. Since fatigue results for flush-ground girth welds in steel pipes have now become
available, it is necessary to re-evaluate the design S-N curve(s) for such welds on the basis of the
relevant experimental data.
For flush-ground joints, internal defects become the weak link in the fatigue performance. The effect of
embedded flaws on fatigue design must also therefore to be re-assessed since the current embedded
flaw acceptance criteria were also based on joints in flat plate rather than girth welds.
This report and the work it describes were funded by the Health and Safety Executive (HSE). Its
contents, including any opinions and/or conclusions expressed, are those of the authors alone and do
not necessarily reflect HSE policy.

HSE BOOKS
© Crown copyright 2003

First published 2003

ISBN 0 7176 2184 7

All rights reserved. No part of this publication may be


reproduced, stored in a retrieval system, or transmitted in
any form or by any means (electronic, mechanical,
photocopying, recording or otherwise) without the prior
written permission of the copyright owner.

Applications for reproduction should be made in writing to:


Licensing Division, Her Majesty's Stationery Office,
St Clements House, 2-16 Colegate, Norwich NR3 1BQ
or by e-mail to hmsolicensing@cabinet-office.x.gsi.gov.uk

ii
CONTENTS
EXECUTIVE SUMMARY v
Background v
Objectives v
Approach v
Conclusions v

1 INTRODUCTION 1
1.1 BACKGROUND 1
1.2 APPROACH 1

2 OBJECTIVES 3

3 FLUSH-GROUND WELDS WITH NO REPORTABLE FLAWS 5


3.1 FLUSH-GROUND GIRTH WELDS 5
3.2 FLUSH-GROUND WELDS IN PLATES 7
3.3 COMPARISON OF FATIGUE CODES REGARDING
FLUSH-GROUND WELDS 7

4 FLUSH-GROUND GIRTH WELDS WITH REPORTABLE FLAWS 9


4.1 FATIGUE DATABASE 9
4.2 EFFECT OF POROSITY 9
4.3 EFFECT OF SLAG INCLUSIONS 9

5 DISCUSSION 11
5.1 STRIP AND FULL-SCALE SPECIMENS 11
5.2 PLATE AND STRIP SPECIMENS 12
5.3 RECOMMENDED S-N CURVE FOR FLUSH-GROUND GIRTH WELDS 12
5.4 EFFECT OF EMBEDDED DEFECTS 13

6 CONCLUSIONS 15

7 RECOMMENDATIONS 17

8 ACKNOWLEDGEMENTS 19

9 REFERENCES 21

TABLES 1 – 6

FIGURES 1 - 11

iii
iv
EXECUTIVE SUMMARY
Background
The fatigue performance of girth welded steel pipes with the weld cap and root flush-ground is
designated as Class C in the BS 7608 code for fatigue design. However, this is not based on
experimental data. Since fatigue results for flush-ground girth welds in steel pipes have now
become available, it is necessary to re-evaluate the design S-N curve(s) for such welds on the
basis of the relevant experimental data.

For flush-ground joints, internal defects become the weak link in the fatigue performance. The
effect of embedded flaws on fatigue design must also therefore to be re-assessed since the
current embedded flaw acceptance criteria were also based on joints in flat plate rather than
girth welds.

Objectives
• To set up a database of S-N values for flush-ground girth welds and then to establish the
best-fit mean and design S-N curves using the new data.

• To evaluate the BS 7608 C curve using these results.

• To compare the BS 7608 C curve with other pertinent fatigue design criteria in other codes
and standards.

• To set up a database of S-N values for flush-ground girth welds with embedded flaws.

• To examine the critical embedded flaw size allowed in such welds for their designated
design S-N curve.

Approach
A comprehensive examination of published works regarding fatigue performance of
flush ground welds with and without reportable flaws was carried out. Discussion and
communications were also held with research workers in this area.

Conclusions
• A database of fatigue test results for flush-ground girth welds has been set up and
statistically analysed.

• There is a controversy in the fatigue performance of full-scale tube specimens. Although


most of the results reviewed strongly support the BS 7608 C curve, limited data, not
reported in full (lacking details of NDT records, failure locations and individual test results)
suggest that Class C can be unsafe. These results may be misleading (e.g. if the welds
contained severe defects) and it is apparent that there is a need for further information. This
requires further effort to establish the missing details and / or more full-scale fatigue testing
of flush-ground girth welds with full details of the welds.

• BS 7608 Class C is comfortably qualified on the basis of fatigue data obtained from strip
specimens cut from girth welded joints. The margin between the test data and the pertinent
curves in other codes, where the effects of seawater in CP and specimen thickness are
considered, is even greater. Thus they are more conservative than BS 7608 Curve C.

v
• The optimistic fatigue performance of small specimens can be partly attributed to their
reduced residual stresses compared to large–scale tube or pipe specimens. Fatigue testing of
such specimens at high stress ratios is required in order to predict the behaviour of actual
girth welds conservatively.

• All S-N curves designated for flush-ground girth welds have been collected and compared.
Although many current fatigue design S-N curves for flush-ground butt-welds have a slope
of m = 3, evidence from the present review indicates that better correlation of the data is
achieved with a shallower slope, m > 3.

• Data from flush-ground plate specimens also support Class C classification. However,
because of some deficiencies such as NDT records, weld quality and misalignment, the
fatigue behaviour of butt-welded plate specimens is not truly representative of that in girth
welds. Therefore they should be used with caution.

• A database of fatigue results for flush-ground girth butt welds containing reportable flaws
has been set up

• The limited database on flush-ground girth welds containing reportable embedded flaws
suggests that the flaw sizes present in the welds which achieved Class C can be detected
using current NDT methods.

vi
1 INTRODUCTION

1.1 BACKGROUND
Flush-grinding of a butt weld is an established method for improving its fatigue performance.
This will eliminate the stress concentration created by the weld profile and remove the inherent
weld toe flaws from which fatigue cracks typically initiate. Consequently a fatigue performance
far superior to that typically assigned to conventional structural welds in the as-welded
condition is expected. However, the current fatigue designs in various standards or codes for
such welds are not based on fatigue test data from flush-ground girth welds, but on data for
joints between flat plates, much of which was obtained many years ago. Furthermore, the
welding procedures, types and consumables might not be representative of those currently used
in offshore structures. However, since data from flush-ground girth welds in steel tubes have
now become available, there is the opportunity to re-evaluate the design S-N curve(s) for such
welds on the basis of relevant experimental data.

To qualify for BS 7608 Class C, the flush-ground welds must be free from significant defects.
However, in the great majority of situations it is found that a fatigue strength higher than Class
D cannot be justified. This is, based mainly on evidence from plate specimen tests, because of
the possible presence of flaws which are too small for reliable detection using the current NDT
methods, but which could be of sufficient size to reduce the fatigue strength of the joint.
Therefore, it is necessary to examine the defect acceptance criteria for flush-ground welds using
data from girth welds, particularly for embedded defects since they become the weak link in the
fatigue performance. In the meantime, the minimum defect sizes need to be assessed against the
detectability of the current NDT methods.

1.2 APPROACH
Although design should ideally be based on fatigue tests on full-scale flush-ground girth welded
tubes or pipes, in order to include the proper effects of size and residual stresses, in reality such
tests can be relatively expensive. Currently such data are very limited in the public domain
(Wirsching et al, 1995, Salama, 1999). Traditionally small specimens have been used to assess
the fatigue performance of large components after appropriate consideration of the differences
between the two. Therefore, test data from strip specimens, extracted from flush-ground girth
welds in steel tubes, which were not available until recently, were collected to re-evaluate the
design S-N curve(s) for such welds. As the current fatigue design curves for girth welded tubes
were traditionally based on plate specimen data, fatigue data from plates with flush-ground
welds were also collected and compared with their counterparts from girth welds. The BS 7608
C curve was used throughout as the reference curve for comparison. An alternative would have
been to have used the 1995 amendment to the HSE Guidance Notes (1990). However, these
were withdrawn in 1998. It is worth noting that the C curve in the HSE Guidance notes in 1990
was the same as that in BS 7608.

To assess the effect of embedded flaws on fatigue performance, fatigue data from flush-ground
welds with reportable defects were also investigated. The test data were compared with
acceptance criteria that are currently used (BS 7910, 1999; ASME, 1993) to determine the
actual maximum acceptable flaw size for a designated classification. The ability of relevant
NDT methods to detect them reliably was also examined.

In this report, the BS 7608 C curve is used as a basis for comparison with the experimental data
in all cases. BS 7608 does not require the application of a specimen thickness correction for
flush ground welds. It also assumes that the fatigue crack growth rate for welds in seawater with
CP is the same as that in air. However, some codes do recognise a specimen thickness effect on

1
fatigue endurance in flush ground welds and an enhanced crack growth rate in seawater with CP
when compared to air. To provide an insight of these effects on the pertinent curves in these
codes, they have been considered for the strip specimens only where the thickness data are
available and some of the experimental data were obtained in seawater with CP. In other cases,
the recommended S-N curves for flush ground welds with the reference thickness are employed.

2
2 OBJECTIVES

• To set up a database of S-N values for flush-ground girth welds and then to establish the
best-fit mean and design S-N curves using the new data.

• To evaluate the BS 7608 C curve using these results.

• To compare the BS 7608 C curve with other pertinent fatigue design criteria in other codes
and standards.

• To set up a database of S-N values for flush-ground girth welds with embedded flaws.

• To examine the critical embedded flaw size allowed in such welds for their designated
design S-N curve.

3
4
3 FLUSH-GROUND WELDS WITH NO REPORTABLE FLAWS

3.1 FLUSH-GROUND GIRTH WELDS


Published data from fatigue tests on full-scale tubes with flush-ground girth welds are very
scarce and only three relevant test series were found (Wirsching et al, 1995; Salama, 1999;
TWI, 2002). Fatigue test results for seven API-5L X60 steel tube specimens with outer diameter
(OD) 610mm and wall thickness (WT) 20mm were reported by Wirsching et al (1995). Details
of the test results, NDT examination records and failure locations were not reported and only the
statistical analysis of fatigue endurances was published. The results reported by Salama (1999)
were from four UOE X65 tube specimens with the same dimensions as those reported by
Wirsching. Again no information on NDT examination or failure locations was reported.

TWI has been involved in fatigue testing of flush-ground welded tubes made of API 5L X60
steel with an OD of 609mm and WT of 21.4mm for tendons, and pipes made of API 5L X80
steel with an OD of 273mm and WT of 12.6mm for risers (TWI, 2002, confidential to the
sponsors). Two tendon specimens containing three welds each were fabricated in the 1G
position, using single sided SAW throughout. The weld root was made onto backing tape. Three
riser specimens containing two welds each were fabricated in the 1G position, using a single
sided GTAW root pass followed by SAW fill and cap. All weld root beads and caps were
ground flush with the pipe surfaces after welding. Comprehensive NDT examinations including
RT, UT and MPI did not reveal any defects. Six full-scale girth welds in the two tendon
specimens and six full-scale girth welds in the three riser specimens were fatigue tested under
tension-tension axial loading at a mean stress of ∼ 175MPa and 125MPa, respectively. All the
tests gave run-outs with no evidence of fatigue cracking in any of the welds.

These results are plotted in Fig.1 together with the BS 7608 C curve. Also shown is the
American AWS Category C1 curve. Strictly speaking, this design curve is intended for
as-welded, not for flush-ground, girth welds; the higher category B is provided for these.
However, for critical applications, the C1 curve has been suggested to be more appropriate for
fatigue design for flush-ground welds (Zettlemoyer 2002; Buitrago and Zettlemoyer, 1999),
although details of the test results supporting this are not available.

As will be seen, the fatigue test results for full-scale specimens do not agree well. Although the
mean curve from Wirsching et al (1995) was reported to be slightly above the BS 7608 C design
curve, the design curve is significantly below the BS 7608 C curve and is even below the AWS
C1 curve after 2x105 cycles. This was because of the large standard deviation in the original
data, due to the small number of the specimens tested. For the tests reported by Salama (1999),
all the individual test results were above the AWS C1 curve, but the endurances of three of the
four tests were below the BS 7608 design C curve. On the other hand, TWI’s tests were all run-
outs and all the data for the twelve welds fall above the C curve. It should be stressed that,
because of the lack of details of the NDT examinations and the stringent requirement on the
elimination of defects to qualify for the Class C curve, the test results from both Wirsching et al
(1995) and Salama (1999) should be taken with caution.

In the early 90’s, a large testing programme, led by TWI (TWI, 1998), was carried out in three
laboratories to investigate the fatigue performance of strip specimens extracted from flush-
ground girth welds in tubes, for applications in the Heidrun TLP. The tendons were made from
12m lengths of 1118mm OD by 38mm WT tubes in steel equivalent to API 5L Grade X70
specification. They were fabricated using a number of welding procedures (TWI, 1998).
Waisted fatigue t est specimens with a minimum cross-section width of around 100mm were

5
extracted from the tubes. All welds were flush-ground both inside and outside before fatigue
testing. The fatigue tests were conducted under constant amplitude axial loading with the stress
ratio R being 0.1 in most cases. The tests were run until failure occurred or to target lives based
on the BS 7608 C curve.

Among the total of 68 specimens, 18 flush-ground strip specimens were tested at 6o C in


seawater with catholic protection (-1050mV). Others were tested at room temperature in air.

In general, establishing the fatigue strengths of the flush-ground butt welds was a challenge. In
the absence of weld toes and any significant embedded flaws, fatigue cracking could initiate at
various locations in the specimens other than the weld. In the event, a test was terminated for
one of four reasons:

• Failure in the weld (11 specimens)


• Failure in the weld but from the edge of the specimen (11 specimens)
• Failure from the machine grips or in parent plate (22 specimens)
• Run-out (i.e. specimen did not fail) (24 specimens)

Thus, only 11 specimens failed in the weld, giving valid endurance data for statistical analysis.
The results were analysed assuming the fatigue endurance, N, of flush-ground welds can be
expressed by the power law equation:

N∆Sm = C [1]

where ∆S is the stress range and C and m are constants. The analysis gave m=4.26, and
C=2.55x1016 for the mean S-N curve with a standard deviation of log N, σ = 0.3127.

The resulting mean and mean-2SD S-N curves are plotted and compared with the C curve in
Fig.2. It can be seen that all the test data and the mean-2SD curve (corresponding to 97.7%
probability of survival) are above the C design curve. Table 1 summarises the S-N curves fitted
to the results for both the strip and tube specimens, as well as the C curve.

Figure 2 also includes the test results of those specimens that did not fail from the weld. Those
specimens, which failed from the specimen edge or away from the weld, were regarded as
'unfailed'. It can be seen that all these data were above the C curve and most of them had
endurances far above the mean curve obtained from the failed specimens. Therefore, the
statistical result from the failed specimens can be regarded as a conservative estimate for strip
specimens. This conservatism was further highlighted by the results of those tests carried out
under catholic protection (CP) environment (TWI, 1998). Although BS 7608 assumes that the
fatigue performance of flush ground welds in seawater with CP is the same as that in air, under
some codes the fatigue damage in such an environment is regarded as being more severe and a
life reduction factor, for example 2.5 in HSE (1990) and DNV (2000), is applied when
compared to the fatigue endurance in air. Furthermore, BS 7608 does not require a thickness
correction for flush ground welds, as mentioned earlier, but other codes do. These effects were
examined for the DNV and HSE codes. The thickness effect in the former is similar to Norsok
(1998) and IIW (1996), while the latter represents a greater penalty on the fatigue endurance of
flush ground welds. The reference thickness and thickness exponent are respectively 25.0mm
and 0.15 for the DNV curve and 16.0mm and 0.3 for the HSE curve. The effects of both
environment (seawater with CP) and specimen thickness on fatigue endurance in DNV and HSE
codes are illustrated in Fig.3, where the fatigue results for the strip specimen, shown in Fig.2,
are included for comparison. It can be seen that the results from the 18 tests in seawater with CP
are far above the 0.76P and C1 curves and none of these specimens failed. By comparing to

6
Fig.2, it can be seen that consideration of thickness and seawater with CP brings the HSE 0.76P
and DNV C1 curves further below the experimental data than the BS 7608 C curve, especially
the HSE 0.76P curve, providing more safety margin. It can also be seen that the fatigue strength
reduction due to specimen thickness up to 38.0mm was very small for the DNV code.

It should be noticed that the fatigue test results from both the full-scale tube and the small-scale
strip specimens exhibited a flatter S-N curve than those obtained for as-welded joints, for which
the slope (m) usually has a value of 3.0 (see for example BS 7608). This is to be expected since
flush grinding introduces a significant crack initiation period, whereas the lives of as-welded
joints consist almost entirely of crack propagation. This difference is reflected in some of the
codes (BS 7608, 1993; ISO, 1999; AWS, 2002), whereas the now withdrawn 1995 amendments
to the HSE Guidance assigned a slope of 3.0 to the curve for flush-ground joints (0.76P).

3.2 FLUSH-GROUND WELDS IN PLATES


As the S-N curve classification of welded plates is the same as that for girth welded tubes, a
review of the fatigue performance of flush-ground butt welded steel plate specimens is useful.
This was conducted recently by Maddox (1997), see Fig.4. Data in the low-cycle fatigue regime
were expressed in terms of the pseudo-elastic stress range (i.e. strain range multiplied by elastic
modulus) as suggested in PD 5500 (2000). As the tensile strength did not have any significant
influence on the fatigue performance in these tests, no distinction was made between the
materials with different tensile strengths.

As will be seen in Fig.4, the plate specimen data are more widely scattered than the girth weld
strip specimen data. The majority of the results qualified for the C curve with only a few
exceptions where the results were only slightly below the C curve. The larger data scatter is to
be expected since the database is much larger than that for strip specimens. However, it might
also be associated with inconsistent welding quality in the plate specimens obtained from many
different sources, in contrast with the comparatively good welding control and stringent NDT
acceptance criteria in girth welding. Consequently, some of the plate specimens might have
contained defects larger than the acceptable limit for girth-welded pipes. Furthermore, the tests
on these plate specimens were carried out at different stress ratios, some even with R<0. As
stress ratio has a strong influence on fatigue endurance (Salama, 1999), tests performed at
different stress ratios can be expected to be scattered. Overall, the results from plate specimens
suggest the applicability of the C curve for flush-ground butt welds.

3.3 COMPARISON OF FATIGUE CODES REGARDING FLUSH-GROUND WELDS


Generally flush-ground butt welds have been recognised as having enhanced fatigue
performance compared with as-welded joints. This is reflected in the different design Standards.
Table 2 summarises the fatigue classification for flush-ground butt welds in the major
international Codes or Standards. A graphic presentation of the corresponding design S-N
curves is shown in Fig.5, where it can be seen that:

• Except for the difference in fatigue limits, the BS 7608 C curve is the same as the ISO
C curve, the DNV C1 curve is the same as the Norsok C1 and the Eurocode3 category 112
curves.
• The Class C curves in BS 7608 and ISO/CD 13819-2 have a slope of 3.5, different from 3.0
adopted by other European codes and IIW. They are more conservative for endurances
below 2x105 cycles but less conservative above this endurance compared to some of the
codes.

7
• As indicated above the 1995 amendments to the HSE Guidance, where the fatigue design
curve for flush-ground welds was designated as the 0.76P curve with a slope of 3.0, was
withdrawn in 1998.

The API designated curve was not included in Fig.5 since API recommends using the AWS’s
guidance (API, 1993). The AWS designation for flush-ground butt welds is Class B which has
a shallow S-N curve slope (m=4.3). As noted earlier, a shallow slope with m>3 would be
expected to be appropriate for flush-ground butt welds as a result of crack initiation occupying
a significant proportion of the fatigue life. Compared to other Standards, AWS B curve is
conservative at fatigue endurances below one million cycles, especially in the regime of high
stress ranges. The AWS C1 curve is also included in Fig.5 for comparison since, as noted
earlier, it has been proposed as a suitable design curve for flush-ground girth welds.
(Zettlemoyer, 2002; Buitrago and Zettlemoyer, 1999).

It must be emphasised that the above comparison is based on the S-N curves at the reference
thickness for each code. Since some codes do consider a thickness effect for flush ground welds
and recommend some degree of penalty, care must be taken when such a comparison is made at
a specimen thickness beyond the reference thickness for a certain code.

8
4 FLUSH-GROUND GIRT H WELDS WITH REPORTABLE FLAWS

4.1 FATIGUE DATABASE


To qualify for the fatigue design curves for flush-ground girth welds, the criteria of defect
acceptance are very strict. Surface breaking and internal planar defects, such as lack of fusion or
lack of penetration, are particularly important since they are much more deleterious to the
fatigue performance of the weld than embedded volumetric defects such as pores or slag
inclusions. In recognition of this, it has been proposed that any indication of their presence
should not be allowed for flush-ground welds in TLP tendons designed to AWS C1 (Buitrago
and Zettlemoyer, 1999).

Recently the effect of porosity and slag inclusions in flush-ground girth welds on fatigue
performance has been investigated (Razmjoo et al 1996, Buitrago and Zettlemoyer, 1999).
Razmjoo et al (1996) carried out fatigue tests on strip specimens containing embedded defects
which were beyond the acceptance limit for production. Buitrago and Zettlemoyer (1999),
however, deliberately introduced embedded defects into welds and investigated their fatigue
performance. These results are summarised in Tables 3, 4 and 5. They are also compared with
the plate specimen data studied by Harrison (1972) in Figs.5, 6 and 7.

4.2 EFFECT OF POROSITY


Fig.6 shows the fatigue test results from both strip and plate specimens containing porosity.
Both porosity density and maximum pore size have been characterised by RT and UT in strip
specimens (Buitrago and Zettlemoyer, 1999; TWI, 1998), but only porosity density was
reported for plate specimens (Harrison, 1972). It will be seen that when porosity density was
less than 4% and the maximum reported individual flaw size was up to 4.8mm, the fatigue
performance could still be qualified as Class C, except in the high stress range/low endurance
(<105 cycles) regime, Fig.6. The results suggest that, in the medium and long fatigue life
regimes, a pore size of around 4mm can be tolerated, a size which can be detected reliably by
RT or UT.

However, when porosity density was increased to 8% many plate specimen test data fell below
the C curve, Fig.7 (Harrison, 1972). This suggests that when pore density is low (below 4%),
the individual pore size plays an important role. When the pore density is high, however, the
fatigue performance is significantly reduced regardless of the individual pore size.

4.3 EFFECT OF SLAG INCLUSIONS


Fatigue test results for flush-ground girth welds in the form of strip specimens (Buitrago and
Zettlemoyer, 1999; TWI, 1998) and butt welds in plates (Harrison, 1972) with slag inclusions
were collected and they are compared in Fig.8. The fatigue data from plates (Harrison, 1972)
were selected with inclusions less than 10mm, but much la rger defects were present in the
specimens reported by Buitrago and Zettlemoyer (1999), see Table 4. It can be seen that the data
from TWI (1998), which had defect lengths between those reported by Harrison (1972) and
Buitrago and Zettlemoyer (1999), qualify for BS 7608 Class C. The data from other tests
(Buitrago and Zettlemoyer, 1999; Harrison, 1972), however, show the effect of embedded
inclusions and many test results fall below the C curve even from the plate specimens. This
suggests that the fatigue life depends not only on the inclusion length as specified in BS 7910,
but also on other factors such as the ligament and the inclusion height.

9
10
5 DISCUSSION

5.1 STRIP AND FULL-SCALE SPECIMENS


When considering fatigue design of tubular joints, results obtained from full-scale girth welded
specimens are highly recommended in order to exclude uncertainties associated with other test
specimens. However, the requirements for the special testing facility, time and cost make this
difficult to achieve. In practice, small-scale specimens have often been used to simulate the
behaviour of actual components. When the test results from these specimens are used for design
purposes, however, care must be taken to ensure they represent the fatigue performance of
large-scale components.

Available fatigue data from flush-ground welds in strip specimens cut from girth welded tubes
confirm the qualification of the BS 7608 C curve. Even those specimens that were originally
rejected due to the presence of unacceptable defects (TWI, 1998) still gave lives consistent with
Class C. In fact, they showed little difference from those failed specimens containing no
reportable defects, see Fig.9. This suggests that the fatigue performance of strip specimens free
from defects can be comfortably classified as Class C. However, some of the results of the
flush-ground girth welded tubes were lower than the corresponding strip specimen results and
could not be qualified for the C curve (Wirsching et al, 1995; Salama, 1999). This can be readily
seen in Fig.10 where a direct comparison of the fatigue performance of strip and full-scale
specimens is shown. Although the data from the above full-scale specimens should be taken
with caution since, as stressed before, no details of NDT, fatigue testing or failure locations
were reported, the difference in fatigue performance between the strip and the full-scale
specimens was not unexpected as the same has been observed in tests on as-welded girth welds
(Maddox and Razmjoo, 1998). In this case, the difference was attributed to the possible effects
of size and residual stress.

With regard to size, as a strip specimen contains only a small proportion of the length of a girth
weld, it is unlikely to contain the most severe defect present and unlikely to introduce the same
level of stress concentration as in some locations in the tube. Thus, it is probable that many of
the strip specimens extracted from a girth weld will exhibit better fatigue performance than the
full-scale girth welded tube.

With regard to residual stress effects, in general it is assumed that welded joints in real
structures contain high tensile residual stresses acting in the most damaging directions, i.e.
transverse to girth welds. It is normally assumed that the residual stress can reach the yield
stress of the material. The effect of the residual stresses is to modify the mean stress of any
applied cyclic stress, giving a relatively high mean stress and a tensile stress range. When a strip
specimen is cut from a pipe, however, the residual stress will be significantly released. Although
the strip and tube specimens were tested at the same nominal stress ratio, the actual stress ratio
experienced might have been different due to a higher residual stress retained in the tube
specimens. This effect can influence both crack initiation and growth, hence the fatigue
performance.

The stress ratio effect on fatigue endurance can be directly seen from a comparison of fatigue
data obtained at different R values, see Fig.11. All these results were obtained from
flush-machined butt welded plate specimens (Oliver and Ritter, 1979). The results show that the
fatigue performance did depend on the stress ratio used. It is also interesting to note that the
results obtained at high stress ratios, R=0.4 and 0.5, were marginally above and below the Class
C curve, respectively. By fatigue testing strip specimens at different stress ratios, Salama (1999)

11
claimed that there was no difference in fatigue performance between the strip and the tube
specimens if the former were tested at a higher mean stress with the maximum stress equal to
the yield stress. It should be noted that residual stresses due to welding were often found to be
widely scattered, from compressive to high tensile stresses even for a single weld (TWI, 2002).
As a consequence, the fatigue performance of girth welds was sometimes sensitive to applied
mean stress and exhibited a large scatter.

Thus, with strip specimens there is a risk of producing non-conservative S-N data. For a better
and conservative correlation between small- and large-scale specimens, the former should be
fatigue tested at a high tensile mean stress or high stress ratio.

With regard to the relative significance of specimen size on residual stress, it is worth noting
that, although it has been generally accepted that the residual stress will be reduced once a strip
specimen is extracted from a girth weld, no direct measurements of the stress release level using
the same component have been reported. It would be very useful if a non-destructive method,
such as neutron diffraction, which can measure residual stresses in thick plates, were used to
measure the residual stresses at the same location both before and after cutting.

Another issue, which could affect the direct comparison of the fatigue results between strip and
full-scale specimens, is the quality of flush-grinding at weld roots. For single sided welds in
tubes, grinding the weld roots is not as easy as for strip specimens. Consequently cracking
might preferentially initiate at the weld root, resulting in a lower fatigue endurance. Thus, to
make the fatigue results of strip specimens representative of large-scale specimens, the same
quality of grinding both inside and outside must be assured.

5.2 PLATE AND STRIP SPECIMENS


In addition to the factors encountered with strip specimens, the application of fatigue data from
plate specimens to large-scale tube components was complicated by the differences from girth
welds. First, no NDT records for these specimens (Maddox, 1997) were reported, which might
explain the large data scatter displayed in Fig.4. As embedded defects can significantly affect
the fatigue endurance of flush-ground welds, the results from these specimens cannot truly
represent girth welds. Secondly, since the fatigue data from these plates were obtained some
years ago, the welds cannot be assumed to be representative of those produced by modern
fabrication processes, particularly for girth welds in pipelines. A special issue is the
misalignment, whic h can significantly affect the local stress distribution around the weld and
hence the fatigue endurance. In view of these factors, it would be prudent to use strip
specimens, extracted from girth welded tubes, rather than from welds made between plates to
simulate the fatigue performance of full-scale tubes, at least in terms of weld quality and
geometry. However, it is interesting to note that, in spite of the differences between the two
kinds of specimens mentioned above, the large database from plate specimens can still
marginally qualify for the C curve.

5.3 RECOMMENDED S-N CURVE FOR FLUSH-GROUND GIRTH WELDS


To qualify for the Class C design curve, detailed NDT examination of welds must be conducted
to ensure that the weld is free from significant defects. Thus, any test data for which the NDT
results are lacking, e.g. the data reported by Wirsching et al (1995) and Salama (1999) on full-
scale fatigue testing, should be taken with caution.

The work carried out by TWI (1998) provided full details of the NDT examination and testing
conditions. All twelve welds qualified with respect to the C curve without failure. The results
therefore strongly support the BS 7608 C curve. Furthermore, all the results from the strip

12
specimens also support the adoption the C curve for flush-ground girth welds even for those
tests undertaken in seawater with catholic protection (Razmjoo et al, 1996).

The statistical analysis of both full-scale and strip specimens reaching failure suggests that a
shallower S-N curve than that recommended in many European guidance documents (DNV,
2000; Norsok, 1998; Eurocode 3, 1992) and IIW (1996) would be appropriate for these welds.

5.4 EFFECT OF EMBEDDED DEFECTS


The defect acceptance criteria in two codes, which might be used for flush-ground girth welds,
are compared in Table 6. One is based on fitness-for-service (BS 7910, 1999) and the other on
fabrication limits (ASME, 2001). It should be noted that the BS 7910 (1999) only provides
defect limits up to quality category Q1 (equivalent to design Class D) on the basis that beyond
this limit NDT cannot be relied upon to detect critical defects. The defect acceptance criteria in
ASME, Section VIII, Division 1 (2001), are those currently used for some tendons (Buitrago
and Zettlemoyer, 1999).

The results obtained from strip specimens cut from girth welded tendons reported by TWI
(1998) provided a direct comparison between specimens containing reportable defects, some of
which were measured, and no reportable defects under the same productio n procedures. A pore
size up to 8mm and a slag inclusion up to 18mm long were reported. The comparable fatigue
endurances of these specimens containing reportable defects with those containing no reportable
defects suggest that these defects can be tolerated, which provides confidence in detecting the
limiting defect sizes using current NDT methods. However, this finding was based on a
comparatively small sample number. More such tests are required to determine the critical
defect size. It was also noticed that when a defect was large (>20mm), the fatigue endurance
began to be reduced below the C curve. Therefore, to be prudent and to determine accurate
defect limits for acceptance (length, ligament, ratio of ligament to thickness etc), a
comprehensive fracture mechanics assessment is required.

13
14
6 CONCLUSIONS

The following conclusions can be drawn on the fatigue performance of flush-ground girth welds
on the basis of the database compiled from test results currently available in the public
literature:

• A database of fatigue test results for flush-ground girth welds has been set up and
statistically analysed.

• There is a controversy in the fatigue performance of full-scale tube specimens. Although


most of the results reviewed strongly support the BS 7608 C curve, limited data, not
reported in full (lacking details of NDT records, failure locations and individual test results)
suggest that Class C can be unsafe. These results may be misleading (e.g. if the welds
contained severe defects), and it is apparent that there is a need for further information. This
requires further effort to establish the missing details and / or more full-scale fatigue testing
of flush-ground girth welds with full details of the welds.

• BS 7608 Class C is comfortably qualified on the basis of fatigue data obtained from strip
specimens cut from girth welded joints. The margin between the test data and the pertinent
curves in other codes, where the effects of seawater in CP and specimen thickness are
considered, is even greater. Thus they are more conservative than BS 7608 Curve C.

• The optimistic fatigue performance of small specimens can be partly attributed to their
reduced residual stresses compared to large–scale tube or pipe specimens. Fatigue testing of
such specimens at high stress ratios is required in order to predict the behaviour of actual
girth welds conservatively.

• All S-N curves designated for flush-ground girth welds have been collected and compared.
Although many current fatigue design S-N curves for flush-ground butt-welds have a slope
of m = 3, evidence from the present review indicates that better correlation of the data is
achieved with a shallower slope, m > 3.

• Data from flush-ground plate specimens also support Class C classification. However,
because of some defic iencies such as NDT records, weld quality and misalignment, the
fatigue behaviour of butt-welded plate specimens is not truly representative of that in girth
welds. Therefore they should be used with caution.

• A database of fatigue results for flush-ground girth butt welds containing reportable flaws
has been set up

• The limited database on flush-ground girth welds containing reportable embedded flaws
suggests that the flaw sizes present in the welds which achieved Class C can be detected
using current NDT methods.

15
16
7 RECOMMENDATIONS

• Fatigue tests on large-scale tubes with flush-ground girth welds with full details of NDT are
required to check the apparently low results obtained from some of the reported large-scale
tests.

• The degree of residual stress relaxation on removing strip specimen from girth welds should
be determined using a non-destructive method such as neutron diffraction.

• If strip specimens are used and the above information is not available, they should be tested
at high stress ratios (R > 0.4) to simulate the effect of high tensile residual stress and so
eliminate the possible dissimilarity from the fatigue behaviour of full-scale components.

17
18
8 ACKNOWLEDGEMENTS

Comments and suggestions made by Dr Nick Zettlemoyer of Exxon Production Research


Company and Professor John Sharp of Cranfield University, U.K., are greatly appreciated.

19
20
9 REFERENCES

1. API RP2A-WSD, 'Recommended Practice for Planning, Designing and Constructing


Fixed Offshore Platforms—Working Stress Design', 20th Edition, American Petroleum
Institute, 1993.

2. ASME, 'Boiler and Pressure Vessels and Code', The American Society of Mechanical
Engineers, Section VIII, Division 1, New York, 2001.

3. AWS D1.1/D1.1M:2002: 'Structural Welding Code - Steel', American Welding Institute,


August 31 2001.

4. BS 7608:1993: 'Fatigue design and assessment of steel structures', BSI, London, UK,
1993.

5. BS 7910:1999: 'Guide on methods for assessing the acceptability of flaws in metallic


structures (incorporating Amendment No. 1)', BSI, London, UK, 1999.

6. Buitrago J and Zettlemoyer N: 'Fatigue of tendon welds with internal defects', Proc. of 18th
International Conference on Offshore Mechanics and Arctic Engineering, OMAE99, Paper
No.: Mat-2001, St. Johns, Newfoundland, Canada, 1999.

7. DD ENV 1993-1-1:1992, 'Design of steel structures—General rules and rules for


buildings', Eurocode 3, 1992.

8. DNV RP C203:2000: 'Fatigue strength analysis', Det Norske Veritas, May 2000.

9. Harrison J D: 'The Basis for An Acceptance Standard for Weld Defects, Part 1: Porosity',
Metal Construction, Vol. 4, pp.99, March 1972.

10. Harrison J D: 'The basis for an acceptance standard for weld defects, Part 2: Slag
inclusions', Metal Construction, Vol. 4, pp.262, July 1972.

11. HSE: 'Guidance on design, construction and certification', 4th Edition, London U.K.,
January 1990 (including amendment No.3 published in Feb. 1995).

12. Hobbacher A: 'Fatigue design of welded joints and components', Recommendations of IIW
Joint Working Group XIII-XV, Abington Publishing, 1996.

13. ISO: 'Draft Revision to ISO 13819-2:1995 Petroleum and Natural Gas Industries –
Offshore Structures – Part 2: Fixed Steel Structures', 18 May 1999.

14. Maddox S J: 'Developments in fatigue design codes and fitness-for-service assessment


methods', Performance of Dynamically Loaded Welded Structures, Proceedings, IIW 50th
Annual Assembly Conference, San Francisco, Ed: S J Maddox, M Prager. Publ: New York,
NY 10017, USA; Welding Research Council, Inc., pp.22-42, 1997.

15. Maddox S J and Razmjoo G R: 'Fatigue performance of large girth welded steel tubes',
Proceedings, 17th International Conference On Offshore Mechanics and Arctic
Engineering, paper No. OMAE 98-2355, 1998.

21
16. NORSOK: 'Design of steel structure', N004, Revision 1, Dec. 1998.

17. Oliver R and Ritter W: 'Catalogue of S-N curves of welded joints in structural steels', DVS,
Dusseldorf, 1979.

18. PD 5500:2000: 'Unfired fusion welded pressure vessels', BSI, January 2000.

19. Razmjoo G R, Hadley I and Crouch S: 'Fatigue of TLP tendon girth weld', Proc. of the 6th
International Offshore and Polar Engineering Conference, Los Angeles, May 1996, pp.190.

20. Salama M M: 'Fatigue design of girth welded pipes and the validity of using strips', Proc.
of 18th International Conference on Offshore Mechanics and Arctic Engineering, OMAE99,
Paper No.: Mat-2003, St. Johns, Newfoundland, Canada, 1999.

21. TWI Report No. 5680/13/98: 'Fatigue performance of flush-ground TLP tendon girth
welds', July 1998.

22. TWI Report No. 5680/26/02: 'Fatigue performance of girth welds made from one side',
May 2002.

23. Wirsching P, Karsan D I and Hanna S Y: 'Fatigue/fracture reliability and maintainability


analysis of the Heidrum TLP tether system', Proc. of 14th International Conference on
Offshore Mechanics and Arctic Engineering, OMAE95, pp.187, Copenhagen, Denmark,
1995.

24. Zettlemoyer N, private communication, 2002.

22
TWI ENDORSEMENT

The report has been reviewed in accordance with TWI policy.

Project Leader............................ TGM/Reviewer ..........................................


(Signature) (Signature)

Print name.................................. Print name .................................................

Secretary/Administrator ..............
(Signature)

Print name..................................
Table 1 Comparison of statistical analysis results of strip and pipe specimens

Strip specimen (TWI, 1998) Tendon specimen BS 7608 C curve


(Wirsching et al, 1995)
m 4.26 3.66 3.5
C, mean curve 2.55x1016 1.33x1014 1.08x1014
C, design curve 6.04x1015 2.84x1013 4.23x1013

Table 2 Details of the fatigue design curves designated for flush-ground girth welds in
different standards. Note the API designation (1993) is the same as AWS (2002)

Standards Parameters Fatigue limit Reference thickness


C m Stress range Endurance (mm)
(MPa) (cycles)
BS 7608- C (1993) 4.23x1013 3.5 78 1.0x107 16
HSE-0.76P (1995) 3.46x1012 3.0 70 1.0x107 16
DNV-C1 (2000) 2.81x1012 3.0 65 1.0x107 25
Norsok-C1 (1998) 2.81x1012 3.0 65 1.0x107 25
Eurocode3-Cat 112 (1992) 2.82x1012 3.0 83 5.0x106 25
ISO-C (1999) 4.26x1012 3.5 41 1.0x108 16
IIW: FAT 125 (1996) 3.91x1012 3.0 92 5.0x106 25
AWS-B (2002) 1.88x1015 4.37 91 5.0x106 not mentioned
AWS-C1 (2002) 9.12x1014 4.33 69 1.0x107 not mentioned
Table 3 Fatigue test and post-mortem results for specimens containing embedded porosity, specimen thickness=20.6mm (Buitrago and
Zettlemoyer, 1999)

Nominal stress Fatigue life Failure site Apparent initiation pore Maximum reported pore size NDE report remarks
range (MPa) (cycles) size (mm) (mm)
200 697,620 Edge N/A 2.4 (<2%) 3 pores, largest: 2.4mm, smallest: 0.8mm
272 335,580 Edge N/A 2.4 (<2%) Cluster, largest: 2.4mm, smallest: 0.8mm
393 17,220 From flaw 3.9 4.8 (<2%) Cluster, largest: 4.8mm, smallest: 2.4mm
212 4,621,260 Edge N/A 1.6 (<2%) Cluster, largest: 2.4mm, smallest: 1.6mm
181 6,454,540 From flaw 1.4 4.8 (<2%) Cluster, largest: 4.8mm, smallest: 1.6mm
272 446,020 Edge N/A 3.2 (<2%) Cluster, largest: 3.2mm, smallest: 0.8mm
200 1,092,780 Edge N/A 2.4 (<2%) Cluster, largest: 2.4mm, smallest: 0.8mm
161 5,217,330 From flaw 3.2 3.2 (2<%) Cluster, largest: 3.2mm, smallest: 1.6mm

Table 4 Fatigue test and post-mortem results for specimens containing embedded linear inclusions, specimen thickness=20.6mm
(Buitrago and Zettlemoyer, 1999)

Nominal stress range Fatigue life (cycles) Failure site Flaw length (mm) Flaw height (mm) Ligament (mm) Maximum reported length (mm)
(MPa)
154 2,413,070 From flaw 28.4 2.8 6.8 25.4
286 35,600 From flaw 24.4 1.6 10.6 22.2
132 12,298,720 From flaw 16.8 1.6 8.7 14.3
117 4,554,370 From flaw 34 1.8 11.7 31.8
198 280,920 From flaw 25.6 1.3 5.6 25.4
286 423,720 From flaw 23.9 1.1 9.5 27.0
198 222,970 From flaw 18.1 1.0 4.4 19.0
286 125,330 Edge N/A N/A N/A 30.0
Table 5 Fatigue test results from specimens containing reportable defects, specimen thickness=38.0mm (TWI, 1998)

Welding procedure Nominal stress Endurance, Details of flaws Failure location


range (MPa) cycles
Two-sided SAW 412 273,640 Slag inclusions. No dimensions reported In the weld
Two-sided SAW 310 1,566,620 Probably slag inclusions. No dimensions reported From specimen edge
Two-sided SAW 359 513,050 Slag inclusions. No dimensions reported From specimen edge
Single -sided SWA with TIG root 343 316,280 Slag inclusions. No dimensions reported From specimen edge
Single -sided SWA with TIG root 489 34,660 Slag inclusions. No dimensions reported From specimen edge
Single -sided SWA with TIG root 405 364,000 Slag inclusions. No dimensions reported From attachment
Single -sided SWA with TIG root 430 101,030 Slag inclusions 6mm long, 8mm ligament From an inclusion
Single -sided SWA with TIG root 423 125,710 Slag inclusions. No dimensions reported In the weld, from inner surface
Single -sided SWA with TIG root 424 143,960 Slag inclusions 4mm long, 5mm ligament From inclusion
Two-sided TIG/MMA 200 3,165,610 Cluster porosity, 8mm long From porosity
Two-sided TIG/MMA 235 498,590 Slag inclusions 16mm long, 8mm ligament From inclusion
Two-sided TIG/MMA 290 333,240 Slag inclusions 18mm long, 6mm ligament From inclusion
Two-sided TIG/MMA 290 577,250 Acceptable NDT indication In the weld, from a tiny subsurface pore
Not reported 400 352,722 No details

Table 6. Defect acceptance criteria for quality category Q1 (BS 7910, 1999) and flush-ground girth weld with thickness limited to 50mm
(ASME, 2001)

Flaw acceptance limit

Standards Porosity, % area on Individual pore size, Slag inclusion length, Undercut, Planer defects
radiograph mm mm (depth/wall thickness)
BS7910-Q1 3.0 Thickness/4 or 6.0 2.5 0.025 Not allowed
ASME VIII Div.1 Not defined 6.3 6.0 No undercut Not allowed
1000

(three welds)

Stress range (MPa)


(three welds)
(two welds) (two welds)

(two welds)

100
Wirsching (610x20mm)
Salama (610x20mm)
TWI (609x21.4mm)
TWI (273x12.6mm)
BS 7608 C curve
AWS C1 curve
10
1.E+04 1.E+05 1.E+06 1.E+07 1.E+08

Endurance, cycles

Fig.1 Comparison of fatigue endurance of full-scale specimens (Wirsching et al, 1995; Salama, 1999; TWI, 2002) with the BS 7608 C and
AWS C1 curves. Solid symbols designate failed tests and open symbols designate run-outs. Note the curve from Wirsching is a design
curve
1000

Nominal stress range (MPa)

100
failed, air
unfailed, air
unfailed, seawater with CP
mean curve (strip specimens)
mean-2SD (strip specimens)
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles

Fig.2 Comparison of the fatigue performance of flush-ground strip specimens cut from girth welds (TWI, 1998) with the BS 7608 C curve
1000

Nominal stress range (MPa)

100
failed, air
unfailed, air
unfailed, seawater with CP
DNV-C1/at ref. thickness
DNV-C1/thickness correction
HSE-0.76P/at ref. thickness
HSE-0.76P/thicknes correction
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles

Fig.3 Comparison of the fatigue performance of flush-ground strip specimens cut from girth welds (TWI, 1998) with the DNV C1 curve and
the HSE 0.76P curve in protected joints in seawater. Specimen thickness effect was also considered as recommended in these codes
10000

Stress range (MPa)


1000

failed (plate specimens)


100
unfailed (plate specimens)
mean (tests)
mean-2SD (tests)
BS7608 C curve
10
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles

Fig.4 Fatigue endurance data for flush-ground butt-welded plate specimens (Maddox, 1997)
1000

Stress range (MPa)

100 BS7608/Class C/1993


HSE/0.76P/1995
DNV/Class C1/2000
Norsok/Class C1/1998
Eurocode3/Category 112/1992
ISO/Class C/1999
IIW/FAT 125/1996
AWS D1.1/Class B/2002
AWS D1.1/Class C1/2002
10
1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08

Endurance, cycles

Fig.5 Comparison of fatigue design curves for flush-ground butt welds from different Standards. Note the API designation (1993) is the
same as AWS (2002)
1000

Stress range (MPa)


3.9mm
3.3mm
1.4mm

3.2mm

100

Harrison
Buitrago & Zettlemoyer
TWI
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles

Fig.6 Fatigue endurance of plate and strip specimens with porosity less than 4%
1000

Stress range (MPa)

100

Harrison

BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles
Fig.7 Fatigue endurance of plate specimens with porosity density up to 8% (Harrison, 1972)
1000

Stress range (MPa)

100

Harrison (failed)
Harrison (unfailed)
TWI
Buitrago & Zettlemoyer
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles
Fig.8 Fatigue performance of flush-ground welded specimens containing slag inclusions. Details of the inclusion sizes and the
corresponding endurance of strip specimens can be found in Tables 3 and 4
1000

Stress range (MPa)

100

failed (no defects)


unfailed, air
failed (with defects)
unfailed (with defects)
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles
Fig.9 Comparison of fatigue performance of strip specimens with and without reportable defects (TWI, 1998)
1000

Stress range (MPa)


(two welds)

(two welds)

(two welds)
(three welds)
100
Wirsching (610x20mm, mean) (three welds)
Salama (610x20mm)
TWI (unfailed)
strip (failed)
strip (unfailed)
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07 1.E+08

Endurance, cycles
Fig.10 Comparison of the fatigue endurance of full-scale and strip specimens with flush-ground girth welds. Solid and circle symbols
denote failed and full-scale specimens, respectively, while the open and square symbols denote run-outs and strip specimens,
respectively
1,000

Stress range (MPa)

100

R=0.5
R=0.4
R=0.25
R=0
BS 7608 C curve
10
1.E+04 1.E+05 1.E+06 1.E+07

Endurance, cycles

Fig.11 Comparison of fatigue performance of welded plates at different stress ratios (Oliver and Ritter, 1979). The BS 7910 C curve is
also included for comparison
Printed and published by the Health and Safety Executive
C30 1/98
Printed and published by the Health and Safety Executive
C1.25 04/03
ISBN 0-7176-2184-7

CRR 090

£20.00 9 780717 621842